Detect Product Faults with a Smart Quality System

In this process we will follow the mentioned guidelines mentioned in the use case understanding and planning:

Understand the data

Generating features

For statistical modelling we need features which needs to be generated by analysing the data. The features which we will consider over here are:

Stats features

  1. Mean value of sensor 1 and sensor 2
  2. Median value of sensor 1 and sensor 2
  3. Standard Deviation value of sensor 1 and sensor 2
  4. Min and max value of sensor 1 and sensor 2

Time features

We will now look at time interval during which these data was generated. We will use rolling mean function over the time window of 5 minutes to compute sensor mean values.

Visualisation

Histogram of quality

Observation: **

Mean plot

Observation: **

Time and Quality

Standard deviation

Observation: **

Modelling

For predicting we are simply doing a classification task with the help of ML algorithms. For this purpose we will use 3 types of algorithms mainly:

  1. Logistic Regression
  2. Decision Trees
  3. Random Forest